pdftex has the very bad habit of picking png first, so we can't
have png images or the pdf version will be horrible. Thus, let's use
gif for the html (which is not picked by pdftex) and a .pdf generated
by our encapsulated postscript sources.

This adds the needed makefile and removes the previous pdf images,
which are regenerated in the next commit. We are also adding new
images where, including the logo used in the cover.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Parts are still to be written, though.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

While the binary control file in sysfs is a neat feature, the
interface to binary attributes was different before 2.6.35. This
allows compiling with no warnings (and no crashes if the feature is
used) by just disabling the attribute at compile time.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

We are getting rid of the "zio-" prefix in all file names.  The prefix
is needed module names (so all single-file devices triggers and
buffers) but not in component files. This simplifies stuff a little
for our command lines and tab completion.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>

Now that both device and trigger/buffer are in objects.c, what
was static should return static.
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

This also removes the ugly __ZIO_INTERNAL__ define we used until now
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

if user specifies only one file, zio-dump assumes it to be a control
file.
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

The -n option tells how many blocks to read from the block source.
The following example will read 3 blocks from zio-zero, channel 0

./zio-dump -n 3 /dev/zio/zzero-0000-0-0-*
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

ZIO creates a binary sysfs attribute named "current-control" within
each channel. This attribute allows the user to read the current
control of any channel.
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubinig@gnudd.com>

The patch 896e0664 misses to assign the zattr for the zio
extended attributes. This patch also improve comments to avoid
similar errors in the future.
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubinig@gnudd.com>

"more_ctrl" is obsoleted by the TLV approach, and I simply forgot to
make this change we agreed about. We have two fields for alarm to
avoid endian issues when checking for errors system-wide (i.e., in all
ZIO devices, including remote ones). The version remains 1.0 as nobody
used the fields in their previous meaning.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

This changes the control block to version 1.0. It adds a TLV header
at the end and renames sockadd_zio to zio_addr (sockaddr_zio must be
defined in user space using the dreaded sa_family_t instead of uint16_t).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Now the open() returns -ENODEV at is should. The bug only
happens if you are not using udev, though.

Original code returned -EBUSY when 0 (== false) was needed. The patch
is bigger than that because it simplifies related code.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

This is an incompatible change for buffer authors. We don't think
there are any out-of-tree buffers, though.  The allocation
function must allocate the control by itself, instead of
receiving it from outside.  The reason why the control was passed
doesn't hold any more, and the upcoming PF_ZIO buffer will
greatly benefit from this.

Also, now alloc_block returns NULL on error, so it can be used directly
by the caller (it used to be ERR_PTR).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

zio-mini is a very simple driver, with no attributes, that can be used
to get accustomed with how ZIO and the ZIO bus work. You can create
as many devices as you want, with as many channels as you want.
Each device has only one input cset, that returns the struct timespec
collected at raw_io() time.

The driver always returns one sample only, even if the trigger
asks for more.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

This replaces the previous allocation bitmask, which only supported
small channel-sets, with the first-fit allocator. Minors are always
taken from the smallest available slot starting from minor 0.

Each cset is allocated separately, but this should not be a problem.

With the zio-mini driver, which is not upstream yet, this works:

     insmod zio-mini.ko  nchan=300 ndev=20

And we get all the devices we need.

    # ls -1 /sys/dev/char/250:* | wc -l
    12000

The patch makes also some other renaming of functions (less underscores
for functions exported by zio-cdev.c to zio-sys.c) and similar stuff
that makes stuff clearer, if I happened to find them during this change.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

This is a library-like set of functions, but they are not in the kernel,
as far as I know, so let's offer them here. We will use them in the
vmalloc buffer, in place of the circular buffer we have now, and for
allocating minor device numbers.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Only useful to diagnose buf-vmalloc, but _there_ it is useful.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

The driver itself is allowed to change nsamples, because it may have
less samples than requested when the trigger fires. This is done by
zio-mini, for example, that will be merged soon.  Therefore, the
number in current_ctrl must be updated, or the previous shorter value
will stick.

Also, move seq_num incrementing earlier, so even if allocation fails
the sequence number is updated (and we need to do alarms too for that).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Cset 4 of zio-loop returns a combined control and data stream to
/dev/zio-loop-4-r-ctrldata . The same will soon happen with
PF_ZIO sockets, so add "-c" to zio-dump to support combined
mode (i.e. control is followed by data).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

The driver supports an output cset that loops to an input cset, and
two output csets of 1 channel each, whose output is returned to two
char devices.  Intended for demonstration and testing.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

This is either one-shot or periodic. For output it uses the timestamp
structure in the control block, for input it doesn't yet.
Output may have some buglets; I'll fix when zio-loop is in place for
diagnostics.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

Just like drivers should update current_ctrl and not block->ctrl,
the same happens to buffers. This only applies to vmalloc, though.

Bug introduced by my commit 6a7378a5:
      zio-sys: delay copying current_ctrl to block->ctrl
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

This removes the annyoing message about the missing attribute.  Simple
driver can choose to not have attributes at all, which simplifies
source code, so the core should better not complain about them.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Federico Vaga <federico.vaga@gmail.com>

This fixes all public names to use zio_ or ZIO_
as a prefix. This introduces incompatibilities for the users,
but was a bad design choice that stuck for too long.
We are going to tag 1.0 soon, so this was unavoidable.